[论文总结] Improving Transferability of Adversarial Examples with Input Diversity 说在前面 CVPR 2018,原文链接:http://arxiv.org/abs/1803.06978 开源代码:https://github.com/cihangxie/DI-2-FGSM 本文作于2021年09月28日。 1. 解决的问题 本文的场景是进行黑盒迁移攻击。在当时,攻击可以分为两类,单步和...
Instead of only using the original images to generate adversarial examples, the proposed method, Diverse Input Iterative Fast Gradient Sign Method (DI2-FGSM), applies random transformations to the input images at each iteration. The generated adversarial examples are much more transferable than those ...